Jaana Dogan

> As in would the global configuration or one that is part of the application be used if both are found. We can have an env variable (e.g. OPENCENSUS_CONFIG) that...

Mutations of context value introduces race conditions. Even though, Go sets no restrictions on value lifecycle, allowing user to mutate them breaks the concept of derived contexts. MutateTags cannot be...

> The concurrency problems with the second approach are no worse than those that come with updating Span attributes This is not the entire picture. Context APIs have some patterns...

I think this is a visualization issue and we cannot do much to make the case better if tracing backends don't allow a pretty name to be visualized as the...

The canonical unit string for unit-less cases we use is "1" but this is deep down documented at http://unitsofmeasure.org/ucum.html and is not very visible. Can we provide a constant for...

LGTM, I don't see any critical problems.

I didn't truly understand why we need to verify go get requests. Don't we instead need to test manually cloning and go get'in from the vanity URL and assert whether...

> What you're suggesting would require that OpenTracing know about every transport/carrier/whatever-you-want-to-call-it that its users want to use. That's not what I am suggesting. Each carrier may implement their own...

> please chime in if you can think of a more concise way of describing the design constraints here... I am trying to summarize some of what was going on...

Pardon me for the late response. I wanted to take the time to read more and understand the decoupling between Tracer's inject/extract, formats and carriers. I was able to understand...